…actually, often they’re both not real: how Weta Digital crafted the Will Smith vs. Will Smith rumble in ‘Gemini Man’
If you’ve read a little about how Weta Digital made a younger version of Will Smith for Ang Lee’s Gemini Man, then you know the overall methodology was to capture Smith with motion capture equipment and then re-create the actor as a fully CG photoreal digital double (in stereo high frame rate, no less).
But how did all that work while shooting the Henry (current age Will Smith) versus Junior (younger Will Smith) shots? I wanted to ask Weta Digital visual effects supervisor Guy Williams exactly how it was done for a specific scene, so we sat down and watched a clip of Henry taking on Junior in the catacombs beneath Budapest.
Here Williams outlined the methodology for shooting and pinpointed several key moments in the fight and the challenges behind creating them. Plus, Weta Digital provided this specific breakdown of the fight scene showing the capture and CG characters in action.
First, a note about fighting style
Guy Williams (visual effects supervisor, Weta Digital): At the beginning of this movie, Ang had this concept of ‘messy fighting’. Ang understands that some performers work in a very choreographed way. The reason why they do certain moves when they fight or even make ‘fighting’ noises is to communicate to each other. It’s almost like a sync pulse, a way to stay in sync with each other so that they know when they hear the sound they’re supposed to be rolling away from the punches coming at them.
There can be two issues here. One, the punches never hit, two, there’s a cadence to it and it can feel very choreographed. A real fight is a lot more frenetic, a lot more violent. You and I have no gentleman’s agreement about when I throw a punch because, if possible, I want to catch you off-guard and land every single punch. So Ang wanted to catch that physical honesty of a real fight. He knew that he couldn’t do that practically because, he said, one previous time he tried to do that – when he tried to break that cadence – on the first take a guy’s nose got broken. So he said, ‘Well…not doing that again!’
Now, one of the benefits of 120 frames per second is that it’s genuine. It’s experiential as opposed to voyeuristic – there’s not very much interpretation. So if you’re going to do that right, you need to be genuine about everything. Your performance has to be genuine. The actors aren’t wearing makeup because that’s not genuine, everything has to be more genuine.
For it to work in this format, we had to come up with an idea called messy fighting, where the premise is that we choreograph an entire fight with the stunt co-ordinator and the fight team. We shoot that entire fight and then we replace it. Meaning, with this scene, the vast majority of it, is fully digital.
Testing a ‘messy fight’
There’s a test we did early on where a fist hits the side of the face, and we looked at all these YouTube videos where people would punch each other and film in super slo-mo. It’s amazing because you see the fist hit the skull and it actually carries forward. The skull starts going sideways but the face is left behind. So all the soft tissue on the faces moved in a very particular way. The eyes anchor, because your skin attaches to the corner of your eyes, and you see the teeth moving, the lips wobble across your teeth, the nose goes sideways and the whole face just rolls around. Then as it recoils it all starts to go the other way.
All that fidelity was important for us. The other thing, and this is where the word ‘messy’ comes from: in a movie fight, I’m going to throw five hundred punches at your face. And I’m going to hit you in the jaw every time. In a real fight, I’m going to miss you, or graze you only. It’s the messiness of a real fight because it’s not choreographed. It’s not designed. So we wanted that to be part of it, too.
What we shot
We basically did the fight multiple times. We did it once on set with the stuntmen. We then did it again a few more times on the mocap stage, first with the stuntmen. So, even though we shot the fight on location, we mocap’d exactly the same fight.
Then we went a step further and had Will Smith – playing the role of Henry – fighting against Junior’s stunt double, Von Simmons. They don’t really hit each other, they don’t go anywhere near that – they’re actually really careful because Will’s got this big camera arm sticking off his face. So instead of throwing punches really close to each other, they’re standing apart shadowboxing a little bit, but it’s enough for Will to act in the moment and give us all the facial capture we need.
Then we do it again, and Travis Parker is now Henry, and Will is now Junior. That means we ultimately filmed the fight four times – once on set and three times on the mocap stage. Now, did we still need to have the stunt guys do a fight if the whole thing is going to be achieved with CG characters? Well, the fight has to be designed at some point. The stunt guys can reach that design cheaper and faster than previs or animation. So I still think it’s best to have the stunt guys do it, and then for us to ‘hijack’ it where necessary. Plus, there are a handful of shots in that scene that are still stunt performers and we just replace their heads. There’s actually a couple of shots in there with real Will Smith, too.
We decided whether to go full-CG on the bodies or use the actors or stunt performers depending on the performance we’re after. If it’s Will Smith’s acting performance that we want, then we keep his performance. If it’s stunt choreography – if it’s fighting – we’re not going to ask Will to do the fighting. We have, instead, these 20 year old fight performers because you need somebody who can do almost in-human level fights. So there we keep their bodies.
There is one moment of serious contact in the fight that happened during during, a moment when the gun butt hits Von in the face. The reason we didn’t have to hijack that with our CG was because Von got clocked with that gun.
When the fight gets really serious, it basically becomes an all-CG shot. We even did a CG simulation on the bone wall in that fifht so that the bones can drop down.
At one point there’s a kick to the head. You’ll notice that Will’s face doesn’t start moving until the foot is on it, and it moves as a function of that, and the body follows. Now, a stunt performer might try to lead with the head, but the body’s already going to be moving in the direction of the head. He has to get up to speed before the foot gets there or else the foot will hit him. But in our case, the body is lagging behind it. The body gets yanked around by the head. You can see that the head wobbles as it comes through. You can actually see this face is distorted. This is what messy fighting is all about.
Adding in extra details
We would also do a little sweat sim to kick sweat off his face as it goes backwards. You actually see the cheeks wobbling on Will Smith because his face is settling after that. At 120 fps, there’s five times as much information and you can actually see the waves going through his face.
The other thing we can do is we’d speed up the action a little bit. We tried to speed it up even more, but it just felt synthetic.
And there were times where we differed from the stunt choreography or mocap to add a punch here or there. We might add a punch here or change a punch if the punch wasn’t working. We would change it to feel a little bit more visceral or authentic. But once again, while we can totally create the fight digitally, it gets very expensive to previs in animation. Previs has a hard time doing this kind of choreography. Previs is more about staging and rough ideas. It’s not about the specifics of punches.
The question is, do I need to shoot the plate or can I just commit to all CG? And in hindsight, we probably didn’t need to shoot half these plates. We could’ve just stuck to CG from the start because it’s actually cheaper. But, in prep and in shooting, it’s really hard to convince people of that truth because they think, if I shoot you a plate, that’s got to be cheaper than going into CG. But of course, if I have to make a clean plate, I have to track it, I have to do all these other things.
My goal in this fight is not to have you go, ‘Man, that is amazing!’ My goal is to have you go, ‘How the hell did they get Will Smith to take those punches?’ I mean, first off, ‘How the hell did they get Will Smith to fight Will Smith?’ But if I can get past that leap of faith, ‘How did any producer agree to let Will Smith be kicked in the face that many times?’